This is a Jupyter Notebook to create a MNIST dataset for Hebrew letters in handwritten manuscripts. The dataset is then used to autodetect Hebrew letters in the manuscripts.
Benjamin Schnabel, 2024
https://www.benjaminschnabel.de
- Python=3.9
- Tensorflow
- Keras
- Pandas
- Numpy
- Matplotlib
- opencv-python
- Pillow
- Jupyter Notebook
Use conda and install Tensorflow. I used tensorflow for Mac M1.
Use github issues.
Just run the Jupyter Notebooks.
- hebrew_letters_dataset.ipynb creates the dataset from the letters in the folder hebrew_letters. The arranged by alphabet.
- The result is the dataset. hebrew_letter_model.keras
- letter_identification.ipynb is used to find the letter in the sample image. Use fine-tuning to adjust the values of the
- The Jupyter Notebook generates a txt file. It contains only the detected line by line.
- There will also be an Excel file generated with the detected letters. The excel file contains the letter, its coordinates and the probability.
- The result looks somewhat like this.
